Synovial Chondromatosis and Soft Tissue Chondroma: Extraosseous Cartilaginous Tumor Deﬁned by FN1 Gene Rearrangement

Modern Pathology (2019) 32:1762–1771 https://doi.org/10.1038/s41379-019-0315-8

ARTICLE

Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor deﬁned by FN1 gene rearrangement

1 1 1 2 1 3 Fernanda Amary ● Luis Perez-Casanova ● Hongtao Ye ● Lucia Cottone ● Anna-Christina Strobl ● Paul Cool ● 2 1 1 1,2 1 1,2 Elena Miranda ● Fitim Berisha ● William Aston ● Maia Rocha ● Paul O’Donnell ● Nischalan Pillay ● 1 4 2 1,2 Roberto Tirabosco ● Daniel Baumhoer ● Edward S. Hookway ● Adrienne M. Flanagan

Received: 31 March 2019 / Revised: 4 June 2019 / Accepted: 4 June 2019 / Published online: 4 July 2019 © The Author(s), under exclusive licence to United States & Canadian Academy of Pathology 2019

Abstract A fusion between ﬁbronectin 1 (FN1) and activin receptor 2A (ACVR2A) has been reported previously in isolated cases of the synovial chondromatosis. To analyze further and validate the ﬁndings, we performed FISH and demonstrated recurrent FN1–ACVR2A rearrangements in synovial chondromatosis (57%), and chondrosarcoma secondary to synovial chondromatosis (75%), showing that FN1 and/or AVCR2A gene rearrangements do not distinguish between benign and malignant synovial chondromatosis. RNA sequencing revealed the presence of the FN1–ACVR2A fusion in several cases

1234567890();,: 1234567890();,: that were negative by FISH suggesting that the true prevalence of this fusion is potentially higher than 57%. In soft tissue chondromas, FN1 alterations were detected by FISH in 50% of cases but no ACVR2A alterations were identified. RNA sequencing identified a fusion involving FN1 and fibroblast growth factor receptor 2 (FGFR2) in the case of soft tissue chondroma and FISH confirmed recurrent involvement of both FGFR1 and FGFR2. These fusions were present in a subset of soft tissue chondromas characterized by grungy calcification, a feature reminiscent of phosphaturic mesenchymal tumor. However, unlike the latter, fibroblast growth factor 23 (FGF23) mRNA expression was not elevated in soft tissue chondromas harboring the FN1–FGFR1 fusion. The mutual exclusivity of ACVR2A rearrangements observed in synovial chondromatosis and FGFR1/2 in soft tissue chondromas suggests these represent separate entities. There have been no reports of malignant soft tissue chondromas, therefore differentiating these lesions will potentially alter clinical management by allowing soft tissue chondromas to be managed more conservatively.

Introduction

Synovial chondromatosis is a rare benign neoplasm associated with the synovium of a joint, tendon sheath, or bursa. The presence of endochondral ossiﬁcation in up to 50% of These authors contributed equally: Fernanda Amary, Luis Perez- cases led to the synonym synovial osteochondromatosis [1]. Casanova Any synovial joint can be involved, but it is more common These authors contributed equally: Edward S Hookway, Adrienne M in larger joints including the knee, hip, and elbow. The Flanagan lesion presents most commonly over 40 years of age and Supplementary information The online version of this article (https:// there is a male predominance (2:1). Histologically, synovial doi.org/10.1038/s41379-019-0315-8) contains supplementary chondromatosis is characterized by multiple hyaline material, which is available to authorized users.

* Edward S. Hookway 2 Cancer Institute, 72 Huntley Street, University College London, [email protected] London WC1E 6BT, UK * Adrienne M. Flanagan 3 a.fl[email protected] The Robert Jones and Agnes Hunt Orthopaedic Hospital NHS Foundation Trust, Oswestry SY10 7AG, UK 4 1 Royal National Orthopaedic Hospital NHS Trust, Bone Tumor Reference Center, Institute of Pathology, University Stanmore, Middlesex HA7 4LP, UK Hospital Basel, Basel, Switzerland Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor defined by FN1. . . 1763 cartilage nodules intimately related to the synovium. These from the RJAH biobank were approved by the National may be embedded within the subsynovial tissue but may Research Ethics Committee of the Health Research also lie in the synovial fluid either attached to the synovium Committee (reference 17/YH/0108; IRAS project identi- or as ‘floating’ bodies. fier: 217446 and licensed by the HTA under number The recent findings of FN1–ACVR2A and ACVR2A–FN1 12073). Ethical approval was also given by the Ethik- in-frame fusions in a case of chondrosarcoma arising sec- kommission bei der Basel (reference 274/12). ondary to synovial chondromatosis and in one of seven All samples were diagnosed by specialist bone tumor cases of synovial chondromatosis [2], indicate that these pathologists (AMF, DB, FA, and RT) according to the lesions represent neoplasms, and that they rarely transform World Health Organization classification 2013 [13, 14]. into secondary chondrosarcoma [3–6]. However, the find- When available, cases were correlated with clinical and ings by Totoki et al., to the best of our knowledge, have not imaging data. All samples were fixed in 10% formal saline, yet been validated. Since peripheral and central chon- decalcified in EDTA or in nitric acid (5%), and processed in drosarcoma, chondroblastoma, phosphaturic mesenchymal paraffin: tissue sections were stained with haematoxylin tumor, and extraskeletal mesenchymal chondrosarcoma are and eosin. among the potential differential diagnoses of synovium- related cartilage tumors, a reliable biomarker could aid Fluorescence in situ hybridization (FISH) significantly in the diagnostic accuracy of these lesions, particularly in needle core biopsies [7–10]. FISH was carried out as described previously [16]. In brief, Soft tissue chondromas occur most commonly in the sections were marked by a specialist histopathologist to hands and feet and share histological features with synovial target tumor-rich area, then deparaffinized and pretreated by chondromatosis, hence it is one of the main differential pressure-cooking following incubation in pepsin solution at diagnoses at these sites. Such lesions are considered to 37 °C for 50 min. Probes were added to tissue sections, represent distinct entities [11–14] and soft tissue chon- denatured at 72 °C, and hybridized overnight at 37 °C. dromas are by definition not related to the synovial lining. Thereafter, the sections were washed and counterstained Nevertheless, even with the use of high-resolution radi- with 4′, 6-diamidino-2-phenylindole (DAPI) and mounted ological imaging in individual cases it can be practically with coverslips. impossible to exclude unequivocally an association with a Probes used for FISH included Agilent SureFISH tendon sheath, a bursa lining, or synovium in the hands and custom-designed FN1 and ACVR2A break-apart probes feet. A specific biomarker would therefore aid in rendering (Agilent, Cheshire, UK) (Table S1) and commercially the correct diagnosis [15]. available break-apart probes for FGF1 (CytoTest, Rock- In this study we set out to determine the prevalence of ville, USA), FGFR1 (Leica Biosystems, Wetzlar, Ger- FN1 and ACVR2A gene rearrangements in a larger set of many), fibroblast growth factor receptor 2 (FGFR2) synovial chondromatosis, and secondary synovial chon- (ZytoVision, Bremerhaven, Germany), and p16/CDKN2A drosarcoma, and to search for recurrent genetic alterations (9p21) (Vysis, Abbott Molecular, IL, USA). A positive in soft tissue chondromas. result was obtained when at least 15% of the nuclei analyzed reveal an aberrant signal on counting a minimum of 50 consecutive nonoverlapping nuclei. Assessment of p16/ Materials and methods CDKN2A FISH was undertaken as previously reported [17]. Samples previously known to have rearrangement for Cases were identified by searching the histopathology FGFR1, FGF1, and FGFR2 by FISH and/or RNA files at the Royal National Orthopaedic Hospital (RNOH), sequencing were employed as controls. UK, The Robert Jones and Agnes Hunt Orthopaedic Hospital (RJAH), and the Basel Bone Tumor Reference RNAscope Centre, Switzerland. The RNOH Biobank was approved by the National Research Ethics Committee of the Health The expression of FGF23 was assessed using the RNA- Research Committee (reference 15/YH/0311: Integrated scope™ assay (Advanced Cell Diagnostics, Hayward, CA, Research Application System (IRAS) project identifier: USA) according to the manufacturer’s instructions. Sections 18309). This specific project was approved by the were hybridized with a probe specific for FGF23 (Hs- National Research Ethics Committee approved UCL/ FGF23, Cat. 557241), and for the endogenous control UCLH Biobank Ethics Committee (specificproject Ubiquitin C (Hs-UBC, Cat. 310041). Hybridization signals reference no. EC17.14). This biobank was licensed by the were amplified and visualized with the RNAscope® 2.5 HD Human Tissue Authority under number 12073. Cases Detection Kit (Brown; Cat. 322360). 1764 F. Amary et al.

Quantitative real time PCR (qPCR) Microsystems, Milton Keynes, UK) using the rabbit monoclonal histone H3F3 K36M mutant antibody (clone RNA was extracted from FFPE sections using the Ambion RM193; RevMAb Biosciences USA, San Francisco, CA, RecoverAll™ Total Nucleic Acid Isolation kit (Ambion, Austin, USA). Leica epitope retrieval 2 solution was used for 20 TX, USA) and reverse transcribed into cDNA with the Invi- min. The antibody was diluted 1:400 [8]. trogen™ SuperScript™ IV First-Strand Synthesis System with random hexamers (Thermo Fisher Scientific, Loughborough, Leicestershire, UK) according to manufacturer’s instructions. Results qPCR was carried out using TaqMan® Gene Expression Assays (Thermo Fisher Scientific): FGF23 (Cat. Hs00221003_m1), and One-hundred and sixteen samples from the same number the endogenous control G6PD (Cat. Hs00166169_m1). For of patients were analyzed by FISH for FN1 and ACVR2A analysis, the means of cycle thresholds (Ct) (n = 2) were cal- gene rearrangement including 90 synovial chon- culated. Based on the ΔCt between the gene and G6PD, the fold dromatoses, seven synovial chondrosarcomas, and 19 soft increase of gene expression was determined. tissue chondromas. Of these, 80 samples (69%) gave informative results (Tables 1 and S2). The percentage of RNA extraction, library preparation, RNA break-apart samples in this study varied from 20 to 80% sequencing of the cells in the cases in which a rearrangement was found. RNA sequencing was performed on cases that were Synovial chondromatosis (n = 90) showed an almost retrieved from the operating theaters immediately after equal gender ratio (1.14M:1F) and presented between the excision. A portion from each of the tumors was snap fro- ages of nine and 94 years of age (mean 48). The knee was zen. Frozen sections were cut and those that were at least the most frequently affected site (47 cases, 52%) followed 60% tumor-rich were homogenized in 1 ml Trizol followed by hand/wrist [11], hip [11], elbow [8], foot/ankle [8], and by extraction using the DirectZol purification kit (Zymo shoulder [5]. Tumor size was available from 57 patients and Research, Irvine, CA, USA) according to the manu- ranged from 4 to 170 mm (mean 49 mm). Of the 40/89 facturer’s instructions. RNA quality was assessed on the patients whose tumors were resected and for whom follow- TapeStation 2200 (Agilent, Santa Clara, CA, USA): a RIN up information was available, 15 (38%) developed a local value of greater than seven was obtained in each case. Poly recurrence between 6 and 236 months (mean 45 months). (A) RNA was selected from 200 ng total RNA using the Three of these 15 patients developed a second recurrence at NEBNext Poly(A) mRNA selection module (NEB, Ipswich, 11, 12, and 90 months following surgery for their first MA, USA), and first-strand libraries constructed with the relapse. The follow-up period of the 25 patients (63%) who NEBNext Ultra Directional RNAseq library preparation kit did not develop a recurrence ranged from 1 to 57 months (NEB) according to the manufacturer’s instructions. (mean 18 months). Libraries were pooled and sequenced with 2 × 101 bp reads Histological assessment revealed multinodular tumors, on a Hiseq 1500 (Illumina, San Diego, CA, USA) in rapid connected to the synovium and composed of chondrocytes run mode. Fusion genes were identified from sequenced often arranged in clusters (clonal pattern of arrangement) RNA with Fusion Catcher using default settings with with mild to moderate variation in cell size. The matrix Bowtie, BLAT, and STAR used as the aligners and frequently demonstrated enchondral ossification and/or GRCh37 as the reference genome [18]. calcification to a varying extent. No histological differences were detected in cases with or without gene rearrangement Immunohistochemistry (Fig. 1). Of the 90 synovial chondromatosis, 58 gave informative Immunohistochemistry was performed in the Leica Bond 3 FISH results for FN1 and/or ACVRA2 gene rearrangements fully automated immunohistochemistry stainer (Leica (Table 1, Table S2), and 57% (33/58) of which revealed a

Table 1 Rearrangement of FN1 and ACVR2A FN1 and ACVR2A rearrangement No rearrangement Any FN1/ACVR2A rearrangement Both FN1 and ACVR2A FN1 only ACVR2A only

Synovial chondromatosis (n = 58) 33 (56.9%) 11 8 14 25 (43.1%) Synovial chondrosarcoma (n = 4) 3 (75%) 1 1 1 1 (25%) Soft tissue chondroma (n = 18) 9 (50%) 0 9 0 9 (50%) Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor deﬁned by FN1. . . 1765

Fig. 1 Synovial chondromatosis of the left knee. Lateral radiograph showing a large calcified mass in the posterior aspect of the left knee, with small calcific foci anteriorly. Photomicrograph (bottom image) showing the classical multinodular pattern of synovial chondromatosis with areas of calcification and endochondral ossification. FISH analysis shows break-apart signals using ACVR2A probe (top right)

Fig. 2 Graphic representation of the gene rearrangement type by site of disease in cases of synovial chondromatosis

Table 2 Characteristics of Case Site Grade Primary/local recurrence FN1 FISH Clinical outcome synovial chondrosarcoma cases ACVR2A CDKN2A

1 Hand III Primary + − Homozygous deletionc Lost to FU 2 Knee II Primarya ++ Homozygous deletion Died of disease 3 Wrist I Local recurrence −− Disomic* Disease free 4 Hip III Local recurrence NI NI NI Died of disease 5 Elbow II Local recurrenceb NI NI NI Lost to FU 6 Hip II Local recurrenceb NI NI NI Lost to FU 7 Knee III Primary − + Copy number gain Disease free Asterisk represents FISH performed on benign synovial chondromatosis; tissue for FISH was not available from the malignant synovial chondrosarcoma aHistological evidence of benign disease abutting chondrosarcoma bTissue at ﬁrst diagnosis not available for review cLoss of two copies of CDKN2A in the presence of one or two centromeric signals rearrangement involving either one or both genes. The FISH results: three of these showed an FN1 and/or ACVRA2 genes involved in the rearrangements by anatomical site are gene rearrangement (Table 2). shown in Fig. 2. In three of the seven cases (cases 1, 2, and 7) synovial Synovial chondrosarcoma was diagnosed in seven chondrosarcoma presented de novo, without a history of con- patients from whose samples were available for analysis ventional synovial chondromatosis. The other (cases 3–6) had a representing 7% (7/97) of the combined benign and history of benign synovial chondromatosis prior to the diag- malignant cohort. Four of the seven cases gave informative nosis of secondary chondrosarcoma (Fig. 3;Table2). 1766 F. Amary et al.

Fig. 3 Photomicrographs of four cases of synovial chondrosarcomas: a, b (Case 2) Interface between the well differentiated and high-grade cellular components; b nuclear polymorphism and an open chromatin appearance in the high-grade component. FISH reveals homozygous deletion of CDKN2A. c, d (Case 3) (c) Multilobular appearance characteristic of synovial chondromatosis, but with (d) large atypical cells exhibiting well-deﬁned borders and an open chromatin pattern. e, f (Case 5). There is signiﬁcant high-grade atypia and loss of lobularity. g, h (Case 6) Tumor lobularity is retained but there is high-grade cytological atypia. FISH for CDKN2A was not available for cases 3, 5, and 6

CDKN2A copy number in synovial chondromatosis Soft tissue chondroma samples were obtained from 19 and secondary chondrosarcoma patients (10F:9M) ranging from 15 to 83 years of age at presentation (mean 56.5 years). Nine tumors were sited in As CDKN2A copy number alterations are commonly seen in the foot, nine in the hand/wrist, and one in the forearm. high-grade conventional chondrosarcoma [17, 19], we Tumor size was available from 14 patients and ranged from investigated CDKN2A copy number by FISH (Table 2). Four 4 to 85 mm (mean 30 mm). Clinical follow-up was available of the seven cases were informative: two showed homo- for eight patients: one tumor was not resected and the other zygous deletion, one copy number gain, and one (case 3) seven have not recurred at the time of analysis (follow-up showed a disomic pattern in the synovial chondromatosis 0–19 months: mean 6 months). component, but the area showing features of chondrosarcoma Histologically, 12/19 soft tissue chondromas showed was not available for analysis (Table 2). Eighteen cases of grungy matrix calcification, five of which exhibited synovial chondromatoses revealed disomy for CNKN2A. chondroblastoma-like features [11] (Fig. 4). Two other Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor defined by FN1. . . 1767

Fig. 4 Soft tissue chondroma of the left third toe. a Dorsoplantar radiograph shows a calcified mass eroding the lateral aspect of the proximal phalanx of the third toe. Sagittal STIR (top) and T1 MR images show a lobular mass containing low signal foci consistent with punctuate calcification, eroding the adjacent bone. Axial STIR MR image (right) showing the hyperintense chondral tumor at the plantar aspect of the third toe (tumor indicated by arrows in all images). b Photomicrograph of a multinodular tumor with grungy-type and c pericellular calcification (chondroblastoma- like). d FN1 gene rearrangement detected by FISH (break-apart signals). FGFR2 gene rearrangement detected by FISH (break-apart signals)

cases showed hyaline matrix without evidence of calcifi- rearrangement was detected were sited in the foot, the cation, and the remaining five cases exhibited a pre- other two occurred in the hand. dominantly myxoid cartilaginous matrix. Two of the cases In view of five cases showing features of with grungy calcification were associated with calcium chondroblastoma-like soft tissue chondroma, we analyzed pyrophosphate deposition disease. these samples for H3.3 K36M expression, an alteration seen Informative FISH results for FN1 and ACVR2A were in 98% of conventional chondroblastomas [8]. None obtained in 18/19 cases. Nine (50%) of the cases exhib- exhibited expression, thereby essentially excluding the ited a FN1 gene rearrangement: all nine belonged to the possibility of H3F3A/B K36M mutation. group with grungy matrix calcification, three of which were classified as the chondroblastoma-like variant RNA sequencing of synovial chondromatosis/ (Fig. 4). None of the cases with a predominantly myxoid chondrosarcoma or hyaline cartilaginous matrix without calcification showed a FN1 gene rearrangement. ACVR2A rearrange- As ACVR2A and/or FN1 rearrangements were detected in ments were not detected in any of the informative cases. 33/58 (57%) cases of synovial chondromatosis and in 9/18 Seven of the nine cases in which an FN1 gene (50%) soft tissue chondromas we next sought other 1768 F. Amary et al.

Fig. 5 Fusion genes identified in synovial chondromatosis and soft fusions identified. d FN1 fusion with chromosome 10 genes FGFR2, tissue chondroma. a FN1–AVCR2A fusion shown in genomic context. CCDC7, and CRTAC1 in genomic context. e Exon map and f resulting b Exon map and c resulting exon composition of FN1–ACVR2A exon composition of FN1–FGFR2 fusion rearrangements by RNA sequencing. An FN1 to ACVR2A FGFR2 (Fig. 5). Five different FN1–FGFR2 transcripts fusion, likely as the result of a 70 MB inversion of chro- were detected in which only the FN1 segment was constant mosome 2, was detected in all three cases of synovial whereas there were different FGFR2 components. These chondromatosis, and the single case of secondary chon- different transcripts likely originated from a single genomic drosarcoma studied (the whole exome of which has been event and were resulted in alternative splicing of the pre- reported previously) [20] (Fig. 5). As FN1 and AVCR2A are mRNA. Only one of the observed FN1–FGFR2 transcripts located on opposite genomic strands, the inversion resulted formed an in-frame fusion that could potentially result in a in a fusion in which both genes are in the same sense as functional chimeric protein. Additional fusions involving each other and in every tumor in which the fusion was FN1 and genes on chromosome 10 were also seen, although present, the predicted protein product was also in-frame. they were detected at much lower expression levels com- The location of the breakpoint within FN1 was different in pared to FN1–FGFR2. A summary of the genetic changes each tumor studied whereas the breakpoint in ACVR2A observed in soft tissue chondroma is presented in supple- always occurred between the second and third exon. The mentary Table 4. chromosomal inversion therefore results in the entire protein-coding length of ACVR2A being appended to a Extended study of FGFR2 rearrangements truncated N-terminus of FN1. In all four cases, a transcript representing the reciprocal arrangement with ACVR2A at We confirmed the FGFR2 rearrangement detected on RNA the 5′ end and FN1 at the 3′ end was also formed, and the sequencing using break-apart FISH probes. We next resulting fusion protein was predicted to be in-frame screened by FISH all cases of synovial chondromatosis, soft (Table S3). tissue chondroma, and synovial chondrosarcoma, in which Previous studies of synovial chondromatosis have break-apart signals were not detected for either ACVR2A revealed a number of karyotypic aberrations including and/or FN1: cases where only one of these genes was abnormalities of chromosome 6 and a (3;14) translocation detected by FISH were also analyzed for the FGFR2 rear- [3]. No direct evidence of these abnormalities was detected rangement. Of the 15 soft tissue chondroma cases tested, by RNA sequencing, however in the absence of a fusion nine had an FN1 rearrangement, and six were wild type for gene being generated, RNA sequencing is not the modality FN1 and ACVR2A. Remaining tissue for analysis was not of choice for investigating genomic structural variations. available from three of the initial cohort tested (leaving 18 informative cases). Six of the nine cases with an FN1 gene A novel recurrent FN1–FGFR2 fusion was identified rearrangement revealed additional FGFR2 rearrangements, in cases of soft tissue chondroma indirectly confirming an FN1–FGFR2 fusion (Table S4). FGFR2 gene rearrangements were not found in any case RNA sequencing of a soft tissue chondroma revealed a of synovial chondromatosis and synovial chondrosarcoma (2;10) translocation in which the 5′ end of FN1 was fused to (Suppl Table S2). Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor defined by FN1. . . 1769

Extended study of FGFR1 rearrangements chondromatosis undergo malignant transformation [21, 22]. The presence of an FN1–ACVR2 fusion did not distinguish Finally, driven by our impression that soft tissue chondroma between benign and malignant disease, but we showed that with grungy calcification shared histological similarities copy number alteration of CDKN2A appears to be a valu- with phosphaturic mesenchymal tumors, known to harbor able predictor of malignancy, similar to conventional car- FN1–FGFR1, and FN1–FGF1 fusion genes [9], we tilaginous tumors [17, 20] where copy number alterations of speculated that both genes represented candidate fusion CDKN2A appear exclusively in chondrosarcomas. genes partners in soft tissue chondromas. We screened the We report for the first time that soft tissue chondromas cases of soft tissue chondromas where tissue was available harbor recurrent FN1–FGFR1 and FN1–FGFR2 gene material (n = 15) for FGF1 and FGFR1 using break-apart fusions demonstrating conclusively that they represent FISH probes: in three cases an FGFR1 gene rearrangement neoplasms. The mutual exclusivityindetectionofFGFR1 was detected. An FN1 gene rearrangement was also iden- and FGFR2 fusion genes in soft tissue chondromas and tified in these three cases by FISH thereby indirectly con- ACVR2A fusions in synovial chondromatosis is supportive firming an FN1–FGFR1 fusion in these cases. An FGF1 of them being distinct pathological entities [13, 14]. The gene rearrangement was not observed in soft tissue chon- ability to distinguish reliably between these two entities is dromas. (Table S4). None of the cases of synovial chon- clinically relevant since soft tissue chondromas have a dromatosis (n = 18) and synovial chondrosarcoma (n = 1) very low risk of local recurrence, and malignant trans- where material was available exhibited either an FGF1 or formation has not yet been described. As a consequence, FGFR1 gene rearrangement. they can be managed more conservatively than synovial chondromatosis. FGF23 mRNA levels in soft tissue chondromas Remarkably we found no difference in the histological appearance between synovial chondromatosis with and In view of an FN1–FGFR1 fusion gene being characteristic without FN1–ACVR2A rearrangements. In contrast, all the of phosphaturic mesenchymal tumor and typically expres- soft tissue chondromas harboring FN1 alterations exhibited sing high levels of FGF23 we investigated this using RNA ‘grungy calcification’,afinding that in some cases is in situ hybridization (RNAscope) and qPCR in eight soft reminiscent of chondroblastomas, and led to the description tissue chondromas cases showing an FN1 break-apart signal of the chondroblastoma-like variant [11]. Phosphaturic of our series. None of the cases studied revealed significant mesenchymal tumors typically also show this particular levels of FGF23 by either test (Supplementary Fig. 1). type of mineralization and are characterized by FN1– FGFR1 fusions in 42% of cases [9], and it was this which prompted us to test our set of soft tissue chondromas for Discussion FGFR1 alterations. However, in contrast to phosphaturic mesenchymal tumors in which FGF23 expression is detec- The current study demonstrates that at least 57% of synovial ted in ~95% of cases [23], we did not detect high levels of chondromatosis cases show a rearrangement involving FN1 FGF23 mRNA expression in the soft tissue chondromas. and/or ACVR2A as detected by FISH. The true prevalence Further study is needed to understand the distinct biological of cases harboring these alterations is likely to be higher on consequences of FN1–FGFR1 fusions in these tumor types the basis that only one of three synovial chondroma/chon- and to explain the expression of FGF23 in PMT but not soft dromatosis cases that revealed an FN1–ACVR2A fusion tissue chondroma. The presence of the same translocation in gene by RNA sequencing demonstrated break-apart signals two tumors of mesenchymal origin is suggestive that there involving both genes. This discrepancy could be explained exists some commonality in the pathogenesis of both enti- by the proximity of both genes on chromosome 2 and the ties. Explanations for the difference between the two tumor limitation of designing appropriate FISH probes to detect types could reside in different cells of origin in which the such a paracentric inversion event. The one case, which was translocation occurred or in the epigenetic composition of not informative for FN1 on FISH highlights the finding that the cells. Irrespective of the mechanism, we consider that it possibly around 30% of cases are noninformative by FISH, would be prudent to measure the serum phosphate levels in demonstrating the need to improve tissue processing of patients whose tumors harbor an FN1 fusion to ensure that calcified samples. any metabolic disturbances can be identified and addressed. We have also confirmed the previously reported occur- Soft tissue chondromas without FN1 rearrangements rence of an FN1 and ACVR2A gene rearrangement in the exhibited a distinctive myxochondroid morphology and malignant variant of synovial chondromatosis [2]. As 7% of might represent a different entity. Indeed, it may be that our cohort was represented by synovial chondrosarcoma our these cases represent a result of a degenerative/reparative data support previous reports that ~10% of synovial process. 1770 F. Amary et al.

In the single case of soft tissue chondroma that was dimerization of ACVR2A inducing activin signaling. In subjected to RNA sequencing, there were several other gene those cases in which FISH did not identify a fusion, it is fusions involving chromosome 10 resulting in predicted in- possible that activation of the FGF pathway has been frame fusions with both FGFR2 and CRTAC1, in addition achieved through other mechanisms or that the disease is to the FN1–FGFR2 fusion. Fusions being formed with two being driven by a mechanism that have yet to be discovered. genes in close proximity to each other on chromosome 10 cannot be explained by a simple single translocation and Acknowledgements This research was funded by a grant to LPC by imply that further genomic rearrangements of the region The Pathological Society of Great Britain and Ireland, the RNOH NHS Trust R&D Office, and Skeletal Action Trust. The Bone Cancer have also occurred. Research Trust provided funding for biobanking to AMF and PC. Of the four synovial chondromatosis/synovial chon- AMF is a NIHR senior investigator and is supported by the National drosarcoma cases subjected to RNA sequencing in our Institute for Health Research, UCLH Biomedical Research Centre, and study, all revealed the presence of fusions involving FN1 the UCL Experimental Cancer Centre. MR is funded by the Tom ACVR2A Prince Cancer Trust. NP is a Cancer Research UK-funded clinician and as previously demonstrated, thereby sup- scientist. We are grateful to the RNOH Musculoskeletal Pathology porting the findings in the larger number of our cases pro- Biobank team for consenting patients and accessing samples, to the bed by FISH. However, it is not possible to state which of UCL Cancer Research UK Cancer Centre core Sequencing and these structurally altered genes drives tumor progression as Pathology facilities. We thank patients for participating in our research and the clinical teams involved in their care. both FN1–ACVR2A and ACVR2A–FN1 are predicted to be FN1 ACVR2A in-frame. Both and have been implicated in Compliance with ethical standards the development of neoplasia. ACVR2A encodes a receptor to Activin A, a protein that is part of the large family of Conflict of interest The authors declare that they have no conflict of transforming growth factor B genes and is involved in interest. regulating multiple cellular processes. ACVR2A is commonly mutated in colorectal carcinoma and as well as Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. hepatocellular and cholangiocarcinoma [24, 25]. FN1 is involved in a number of cell processes including adhesion References and migration, the aberrant regulation of which could contribute to a neoplastic process. In addition to phospha- FN1 1. Walker EA, Murphey MD, Fetsch JF. Imaging characteristics of turic mesenchymal tumors discussed above, fusions tenosynovial and bursal chondromatosis. Skelet Radiol. with ALK have been described in gastrointestinal leiomyo- 2011;40:317–25. mas [26], and EGF fusions in calcifying aponeurotic 2. Totoki Y, Yoshida A, Hosoda F, Nakamura H, Hama N, Ogura K, fibromas [27]. et al. Unique mutation portraits and frequent COL2A1 gene alteration in chondrosarcoma. Genome Res. 2014;24:1411–20. In terms of explaining the mechanism of action of the 3. Buddingh EP, Krallman P, Neff JR, Nelson M, Liu J, Bridge JA. fusion genes in synovial chondromatosis and soft tissue Chromosome 6 abnormalities are recurrent in synovial chon- chondromas, it is noteworthy that FN1 is found as the 5′ dromatosis. Cancer Genet Cytogenet. 2003;140:18–22. fusion partner of both ACVR2A and FGFR2, and that the 3′ 4. Mertens F, Jonsson K, Willen H, Rydholm A, Kreicbergs A, ′ Eriksson L, et al. Chromosome rearrangements in synovial fusion partners have not been found with other 5 partners. chondromatous lesions. Br J Cancer. 1996;74:251–4. This implies that the FN1 component of the fusion plays an 5. Sciot R, Dal Cin P, Bellemans J, Samson I, Van den Berghe H, important biological role in the pathogenesis of these Van Damme B. Synovial chondromatosis: clonal chromosome changes provide further evidence for a neoplastic disorder. tumors and is not acting simply through driving the – ′ Virchows Arch. 1998;433:189 91. expression of the 3 partner through an active promoter. 6. Tallini G, Dorfman H, Brys P, Dal Cin P, De Wever I, Fletcher The FN1–FGFR2 fusion identified by RNA sequencing in CD, et al. Correlation between clinicopathological features and the soft tissue chondroma in our study reveals a similar karyotype in 100 cartilaginous and chordoid tumours. A report from the chromosomes and morphology (CHAMP) collaborative domain structure of the resultant proteins to that found – FN1–FGFR1 study group. J Pathol. 2002;196:194 203. previously involving the fusion in phospha- 7. Amary MF, Bacsi K, Maggiani F, Damato S, Halai D, Berisha F, turic mesenchymal tumor [9]. In both cases, the N-terminal et al. IDH1 and IDH2 mutations are frequent events in central region of FN1 is attached to a truncated extracellular region chondrosarcoma and central and periosteal chondromas but not in – of an FGF receptor leaving the transmembrane and intra- other mesenchymal tumours. J Pathol. 2011;224:334 43. 8. Amary MF, Berisha F, Mozela R, Gibbons R, Guttridge A, cellular domains intact. As previously hypothesized by Lee O’Donnell P, et al. The H3F3 K36M mutant antibody is a sen- et al., the presence of the N-terminal region of FN1 may aid sitive and specific marker for the diagnosis of chondroblastoma. the dimerization of the fused FGF receptor through the Histopathology. 2016;69:121–7. fi 9. Lee JC, Su SY, Changou CA, Yang RS, Tsai KS, Collins MT, bronectin domain leading to signal activation [9]. This – – FN1– et al. Characterization of FN1 FGFR1 and novel FN1 FGF1 same mechanism may explain the action of the fusion genes in a large series of phosphaturic mesenchymal ACVR2A fusion, namely that FN1 may facilitate tumors. Mod Pathol. 2016;29:1335–46. Synovial chondromatosis and soft tissue chondroma: extraosseous cartilaginous tumor defined by FN1. . . 1771

10. Wang L, Motoi T, Khanin R, Olshen A, Mertens F, Bridge J, et al. 18. Nicorici D, Satalan M, Edgren H, Kangaspeska S, Murumagi A, Identification of a novel, recurrent HEY1-NCOA2 fusion in Kallioniemi O, et al. FusionCatcher—a tool for finding somatic mesenchymal chondrosarcoma based on a genome-wide screen of fusion genes in paired-end RNA-sequencing data. bioRxiv exon-level expression data. Genes Chromosomes Cancer. 2012; 2014:011650. 51:127–39. 19. Schrage YM, Lam S, Jochemsen AG, Cleton-Jansen AM, Tami- 11. Cates JM, Rosenberg AE, O’Connell JX, Nielsen GP. niau AH, Hogendoorn PC, et al. Central chondrosarcoma pro- Chondroblastoma-like chondroma of soft tissue: an under- gression is associated with pRb pathway alterations: CDK4 down- recognized variant and its differential diagnosis. Am J Surg regulation and p16 overexpression inhibit cell growth in vitro. J Pathol. 2001;25:661–6. Cell Mol Med. 2009;13:2843–52. 12. Lichtenstein L, Goldman RL. Cartilage tumors in soft tissues, 20. Tarpey PS, Behjati S, Cooke SL, Van Loo P, Wedge DC, Pillay particularly in the hand and foot. Cancer. 1964;17:1203–8. N, et al. Frequent mutation of the major cartilage collagen gene 13. Sciot R, Bridge JA. Synovial chondromatosis. In: Fletcher CDM, COL2A1 in chondrosarcoma. Nat Genet. 2013;45:923–6. Bridge JA, Hogendoorn PC, editors. WHO classification of 21. Bertoni F, Unni KK, Beabout JW, Sim FH. Chondrosarcomas of tumours of soft tissue and bone. Lyon, France: IARC Press; 2013. the synovium. Cancer. 1991;67:155–62. p. 261. 22. Biazzo A, Confalonieri N. Synovial chondrosarcoma. Ann Transl 14. Rosenberg AE, Mandahl N. Soft-tissue chondroma. In: Fletcher Med. 2016;4:280. CDM, Bridge JA, Hogendoorn PC, editors. WHO classification of 23. Carter JM, Caron BL, Dogan A, Folpe AL. A novel chromogenic tumours of soft tissue and bone. Lyon, France: IARC Press; 2013. in situ hybridization assay for FGF23 mRNA in phosphaturic p. 160–1. mesenchymal tumors. Am J Surg Pathol. 2015;39:75–83. 15. Damato S, Alorjani M, Bonar F, McCarthy SW, Cannon SR, 24. Yuza K, Nagahashi M, Watanabe S, Takabe K, Wakai T. O’Donnell P, et al. IDH1 mutations are not found in cartilaginous Hypermutation and microsatellite instability in gastrointestinal tumours other than central and periosteal chondrosarcomas and cancers. Oncotarget. 2017;8:112103–15. enchondromas. Histopathology. 2012;60:363–5. 25. Wang A, Wu L, Lin J, Han L, Bian J, Wu Y, et al. Whole-exome 16. Amary MF, Berisha F, Bernardi Fdel C, Herbert A, James M, sequencing reveals the origin and evolution of hepato- Reis-Filho JS, et al. Detection of SS18-SSX fusion transcripts in cholangiocarcinoma. Nat Commun. 2018;9:894. formalin-fixed paraffin-embedded neoplasms: analysis of con- 26. Panagopoulos I, Gorunova L, Lund-Iversen M, Lobmaier I, ventional RT-PCR, qRT-PCR and dual color FISH as diagnostic Bjerkehagen B, Heim S. Recurrent fusion of the genes FN1 and tools for synovial sarcoma. Mod Pathol. 2007;20:482–96. ALK in gastrointestinal leiomyomas. Mod Pathol. 2016;29: 17. Amary MF, Ye H, Forbes G, Damato S, Maggiani F, Pollock R, 1415–23. et al. Isocitrate dehydrogenase 1 mutations (IDH1) and p16/ 27. Puls F, Hofvander J, Magnusson L, Nilsson J, Haywood E, CDKN2A copy number change in conventional chondrosarcomas. Sumathi VP, et al. FN1-EGF gene fusions are recurrent in calci- Virchows Arch. 2015;466:217–22. fying aponeurotic fibroma. J Pathol. 2016;238:502–7.